1 / 27

Designing Extensible IP Router Software

Designing Extensible IP Router Software. Mark Handley (UCL), Eddie Kohler (UCLA), Atanu Ghosh, Orion Hodson, Pavlin Radoslavov (ICSI). Stalled Evolution in the Internet Core. Stability matters above all else . ISPs can’t afford routing problems.

jael
Télécharger la présentation

Designing Extensible IP Router Software

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Designing Extensible IP Router Software Mark Handley (UCL), Eddie Kohler (UCLA), Atanu Ghosh, Orion Hodson, Pavlin Radoslavov (ICSI)

  2. Stalled Evolution in the Internet Core Stability matters above all else. • ISPs can’t afford routing problems. • Won’t run anything experimental on production routers for fear of affecting production traffic. Building stable routing protocol implementations is really hard. • Big router vendors don’t have a lot to gain from innovation, because ISPs can’t afford to run experimental code. • Startups can’t help because of the closed nature of the software market for IP routers. Important routing problems remain unaddressed.

  3. Extensible Router Control Plane Software • Extensibility could solve this problem • Allow experimental deployment of new routing protocols • Enable router application market • Extensible forwarding planes exist • Network Processors, FPGAs, software (Click, Scout, ...) • But not control planes: why? • Router software demands make extensibility hard • Massive scale • Stability requirement • Tight coupling between functions • Routing protocols themselves not designed for extensibility

  4. Fast Convergence Routing protocol implementations have often been scanner-based. • Periodically a scanner runs to accumulate changes, update the forwarding table, notify neighbors, etc. • Easy to implement. • Low CPU utilization. • Poor route convergence properties. Fast convergence is now a priority. • Event-driven router implementations are needed to respond to change as quickly as possible. • Events processed to completion. • Explicit dependency tracking. • Harder to implement, especially in an extensible way.

  5. XORP: eXtensible Open Router Platform Open source router software suite, designed from the outset with extensibility in mind. • Main core unicast and multicast routing protocols. • Dynamically extensible management framework, command line interface, and SNMP. • Event-driven multi-process architecture. • Scriptable inter-process communication mechanism. • 560,000 lines of C++

  6. Contributions • Staged design for BGP, RIB. • XRL IPC mechanism. • Dynamically extensible command-line interface and router management software. • Safe route iterators (see paper). • First fully extensible, event-driven, open-source routing protocol suite: www.xorp.org.

  7. XORP Processes Multi-process architecture, providing isolation boundaries between separate functional elements. Flexible IPC interface between modules

  8. Routing process design 2. Extensible management framework Outline of this talk 3. Performance results

  9. Conventional router implementation

  10. Implementing for Extensibility • Tightly coupled architectures perform well, but are extremely hard to change without understanding how all the features interact. • Need an architecture that permits future extension, while minimizing the need to understand all the other possible extensions that might have been added. • We chose a data-flow architecture. • Routing tables are composed of dynamic processes through which routes flow. • Each stage implements a common simple interface.

  11. BGP Staged Architecture

  12. tree of routes • Unmodified routes stored at ingress • Changes in downstream modules (filters, nexthop state, etc) handled by PeerIn pushing the routes again. Messages add_route delete_route Filter Bank Peer In lookup_route

  13. BGP Staged Architecture

  14. Decomposing BGP Decision

  15. Dynamic Stages PeerIn is ready for peering to come back up Peering Went Down! Deletion Stage does background deletion Take Entire Route Table from PeerIn Problem 1: deleting 150,000 routes takes a long time. Problem 2: peering may come up again while we’re still deleting the routes

  16. RIB Structure Routing protocols can register interest in tracking changes to specific routes.

  17. transport: eg x-tcp, x-udp, kill, finder module name: eg bgp, rip, ospf, fea interface name: eg bgp, vif manager typed parameters to method method name: set_bgp_as, delete_route, etc Inter-process Communication XORP Resource Locators (XRLs): • URL-like unified structure for inter-process communication: • Eg: finder://bgp/bgp/1.0/set_bgp_as?as:u32=1777 • Finder resolves to a concrete method instance, instantiates transport, and performs access control. xtcp://192.1.2.3:8765/bgp/1.0/set_bgp_as?as:u32=1777

  18. Inter-process Communication • XRLs support extensibility by allowing “non-native” mechanisms to be accessed by unmodified XORP processes. • Add new XRL protocol families: eg kill, SNMP • ASCII canonical representation means XRL can be scripted from python, perl, bash, etc. • XORP test harnesses built this way. • Router manager is dynamically extensible using declarative ASCII template files linking configuration state to XRL to instantiate that configuration.

  19. Router Manager template filesMap Juniper-style configuration state to XRLs protocols ospf { router-id: 128.16.64.1 area 128.16.0.1 { interface xl0 { hello-interval: 30 } } } protocols.ospf { area.ADDR { interface.IFNAME { hello-interval { %set: xrl "ospfd/set_interface_param ? area_id:ipv4=ADDR & interface:txt=IFNAME & ospf_if_hello_interval:i32=VALUE"; } } } }

  20. Evaluation • Was performance compromised for extensibility?

  21. Performance: Time from received by BGP to installed in kernel

  22. Performance:Where is the time spent?

  23. Performance: Time from received by BGP until new route chosen and sent to BGP peer

  24. Summary • Designing for extensibility is difficult • Needs to be a priority from day one. • Anecdotal evidence shows XORP to be easy to extend (once you’ve learned the basics) • Performance always at odds with modularity and extensibility. • For routing software, scalability matters most. • Modern CPUs change the landscape, and make better solutions possible. • Only time will tell if XORP is adopted widely enough to change the way Internet protocols are developed, tested, and deployed.

  25. Thanks XORP developers (past and present): • Adam Barr, Fred Bauer, Andrea Bittau, Javier Cardona, Atanu Ghosh, Adam Greenhalgh, Tim Griffin, Mark Handley, Orion Hodson, Eddie Kohler, Luigi Rizzo, Bruce M. Simpson, Pavlin Radoslavov, Marko Zec Financial support for XORP has come from: • ICSI • Intel Corporation • National Science Foundation • DARPA • Microsoft Corporation • The Royal Society.

  26. Registering Interest in Routes Routes in RIB: 128.16.0.0/16 128.16.0.0/18 128.16.0.0/18 128.16.128.0/17 128.16.128.0/17 128.16.192.0/18 BGP interested in 128.16.32.1 interested in 128.16.160.1 128.16.0.0/18 128.16.128.0/18

  27. Libxorp • Libxorp contains basic data structures that can be used by XORP processes. • Example: timers, selectors, route trees. • One detail: safe route iterators • Background task like a deletion stage needs to keep track of where it was in walking the route tree. • New events can cause routes to be deleted. • Solution: safe route iterators, combined with reference counted data structures, ensure that the iterator will never be left invalid.

More Related